An Algorithm for Constructing Feature Relations between the Classes in the Trading Set

In this paper, we propose a new approach for determining the feature relations between classes to solving a problem of classification on based (precedents of) training set. The main Idea of this approach is the selection of best features for object classification in the condition of increasing the number of classes and features in the training set. For this, pay attention to two criteria: increase the recognition quality and reduce the number of calculations by removing non-informative features.